U.S. patent number 7,167,745 [Application Number 10/816,549] was granted by the patent office on 2007-01-23 for methods for quantifying the morphology and amplitude of cardiac action potential alternans.
This patent grant is currently assigned to Quinton Cardiology, Inc.. Invention is credited to David Milton Hadley, Mustafa Hikmet Sagiroglu.
United States Patent |
7,167,745 |
Hadley , et al. |
January 23, 2007 |
**Please see images for:
( Certificate of Correction ) ** |
Methods for quantifying the morphology and amplitude of cardiac
action potential alternans
Abstract
Methods and apparatus for determining T-wave alternan signatures
(i.e., morphology and polarity) derived from a physiologic signal
representative of a subject's heart activity; assessing changes in
the myocardium Action Potential ("AP") through analysis of the
alternan signature derived from a physiologic signal representative
of a subject's heart activity; and/or assessing spatial
disassociation of alternan characteristics that are likely
associated with the initiation of re-entrant arrhythmias.
Inventors: |
Hadley; David Milton
(Woodinville, WA), Sagiroglu; Mustafa Hikmet (Bellevue,
WA) |
Assignee: |
Quinton Cardiology, Inc.
(Bothell, WA)
|
Family
ID: |
34964966 |
Appl.
No.: |
10/816,549 |
Filed: |
March 30, 2004 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20050222512 A1 |
Oct 6, 2005 |
|
Current U.S.
Class: |
600/515 |
Current CPC
Class: |
A61B
5/349 (20210101); A61B 5/7239 (20130101) |
Current International
Class: |
A61B
5/04 (20060101) |
Field of
Search: |
;600/508,509,517,515 |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
Ackerman, et al., "Ion Channels--Basic Science and Clinical
Disease," New England Journal of Medicine, vol. 336 (22), pp.
1575-1586, 1977. cited by other .
Adam, et al., "Estimation of Ventricular Vulnerability to
Fibrillation Through T-Wave Time Series Analysis," Computers in
Cardiology, pp. 307-310, Sep. 1981. cited by other .
Adam, et al., "Fluctations in T-Wave Morphology and Susceptibilty
to Ventricular Fibrillation," Journal of Electrocardiology, vol.
17(3), pp. 209-218, 1984. cited by other .
Adam, et al., "Ventricular Fibrillation and Fluctutations in the
Magnitude of the Repolarization Vector," Computers in Cardiology,
pp. 241-244, 1982. cited by other .
Carson, et al., "Characterisation of unipolar waveform alternation
in acutely ischaemic porcine myocardium," Cardiovascular Research,
vol. 20, pp. 521-527, 1986. cited by other .
Chinushi, et al., "Electrophysiological Basis of Arrhythmogenicity
of QT/T Alternans in the Long-QT Syndrome--Tridimensional Analysis
of the Kinetics of Cardiac Repolarization," Circulation Research,
vol. 83 (6), pp. 614-628, Sep. 21, 1998. cited by other .
CINCA, et al., "Mechanism and Time Course of the Early Electrical
Changes During Acute Coronary Artery Occlusion--An Attempt to
Correlate the Early ECG Changes in Man to the Cellular
Electrophysiology in the Pig," Chest, vol. 77, pp. 499-505, Apr.
1980. cited by other .
Coetzee, et al., "Effects of thiol-modifying agents on K.sub.ATP
channels in guinea pig ventricular cells," American Journal of
Physiology, vol. 38, pp. H1625-H1633, 1995. cited by other .
Coronel, et al., "Reperfusion arrhythmias in isolated perfused pig
hearts--Inhomogeneities in extracellular potassium, ST and TQ
potentials, and transmembrane action potentials," Circulation
Research, vol. 71 (5), pp. 1131-1142, Nov. 1992. cited by other
.
Demidowich, et al., "Electrical altemans of the ST segment in
non-Prinzmetal's angina," PACE, vol. 3, pp. 733-736, Nov.-Dec.
1980. cited by other .
Di Bernardo, et al., "Effect of changes in heart rate and in action
potential duration on the electrocardiogram T wave shape," Abstract
only, Physiol Meas. vol. 23 (2), pp. 355-364, May 2002. cited by
other .
Duckett, et al., "Modeling the Dynamics of Cardiac Action
Potentials," Physical Review Letters, vol. 84 (4), pp. 884-887,
Jul. 24, 2000. cited by other .
Gima, et al., "Ionic Current Basic of Electrocardiographic
Waveforms--A Model Study," Circulation Research, vol. 90, pp.
889-896, May 2002. cited by other .
Han, "Ventricular vulnerability during acute coronary occlusion,"
American Journal of Cardiology, vol. 24, pp. 857-864, Dec. 1969.
cited by other .
Han, et al., "Temporal dispersion of recovery of excitability in
atrium and ventricle as a function of heart rate," American Heart
Journal, vol. 71 (4), pp. 481-487, Apr. 1966. cited by other .
Hashimoto, et al., "Effects of calcium antagonists on the
electrical alternans of the ST segment and on associated mechanical
alternans during acute coronary occlusion in dogs," Circulation,
vol. 68 (3): 667-672, Sep. 1983. cited by other .
Hashimoto, et al., "Effect of the ventricular premature beat on the
alternation of the repolarization phase in ischemic myocardium
during acute coronary occlusion in dogs," Abstract only, Journal of
Electrocardiology, vol. 17 (3), pp. 229-238, Jul. 1984. cited by
other .
Hellerstein, et al., "Electrical alternation in experimental
coronary artery occlusion," American Journal of Physiology, vol.
160, pp. 366-374, Feb. 1950. cited by other .
Kass, et al., "Channel structure and drug-induced cardiac
arrhythmias," PNAS, vol. 97 (22), pp. 11683-11684, Oct. 24, 2000.
cited by other .
Ka{hacek over (z)}i et al., "Ion Channels and Drug
Development--Focus on Potassium Channels and Their Modulators,"
Medicine and Biology, vol. 6 (1), pp. 23-30, 1999. cited by other
.
Kleinfeld, et al., "Alternans of the ST Segment in Prinzmetal's
Angina," Circulation, vol. 55 (4), pp. 574-577, Apr. 1977. cited by
other .
Kleinfeld, et al., "Electrical alternans of components of action
potential," American Heart Journal, vol. 75 (4), pp. 528-530, Apr.
1968. cited by other .
Konta, et al., "Significance of discordant ST alternans in
ventricular fibrillation," Circulation, vol. 82 (6), pp. 2185-2189,
Dec. 1990. cited by other .
Kubota, et al., "Role of ATP-Sensitive K.sup.+Channel of ECG ST
Segment During a Bout of Myocardial Ischemia--A Study of Epicardial
Mapping in Dogs," Circulation, vol. 88 (4, Part 1), pp. 1845-1851,
Oct. 1993. cited by other .
Kurz, et al., "Ischaemia induced alternans of action potential
duration in the intact-heart: dependence on coronary flow, preload
and cycle length," European Heart Journal, vol. 14, pp. 1410-1420,
1993. cited by other .
Lukas, et al., "Differences in the electrophysiological response of
canine ventricular epicardium and endocardium to ischemia: Role of
the transient outward current," Circulation, vol. 88 (6), pp.
2903-2915, Dec. 1993. cited by other .
Moody, et al., "Clinical Validation of the ECG-Derived Respiration
(EDR) Technique," Computers in Cardiology, pp. 507-510, 1986. cited
by other .
Nakashima, et al., "Experimental studies and clinical report on the
electrical alternans of ST segment during myocardial ischemia,"
Japanese Heart Journal, vol. 19 (3) pp. 396-408, May 1978. cited by
other .
Nearing, et al., "Dynamic Tracking of Cardiac Vulnerability by
Complex Demodulation of the T Wave," Science, vol. 252, pp.
437-440, Apr. 1991. cited by other .
Nearing, et al., "Modified moving average analysis of T-wave
alternans to predict ventricular fibrillation with high accuracy,"
Journal of Applied Physiology, vol. 92, pp. 541-549, Feb. 2002.
cited by other .
Nearing, et al., "Tracking States of Heightened Cardiac Electrical
Instability by Computing Interlead Heterogeneity of T-Wave
Morphology Using Second Central Moment Analysis," J Appl Physiol,
vol. 95, pp. 2265-2272, Dec. 2003., 41 pages (First published Aug.
1, 2003; 10.1152/japplphysiol.00623.2003). cited by other .
Pastore, et al., "Mechanism Linking T-Wave Alternans to the Genesis
of Cardiac Fibrillation," Circulation, vol. 99, pp. 1385-1394, Mar.
1999. cited by other .
Raeder, et al., "Alternating Morphology of the QRST Complex
Preceding Sudden Death," New England Journal of Medicine, vol. 326
(4), pp. 271-272, Jan. 23, 1992. cited by other .
Ring, et al., "Exercise-Induced ST Segment Alternans," American
Heart Journal, vol. 111 (5), pp. 1009-1011, May 1986. cited by
other .
Roden, et al., "Cardiac Ion Channels," Annual Review Physiology,
vol. 64, pp. 431-475, 2002. cited by other .
Salerno, et al., "Ventricular arrhythmias during acute myocardial
ischaemia in man. The role and significance of R-ST-T alternans and
the prevention of ischaemic sudden death by medical treatment,"
European Heart Journal, vol. 7 Suppl A, pp. 63-75, 1986. cited by
other .
Schram, et al., "Differential Distribution of Cardiac Ion Channel
Expression as a Basis for Regional Specialization in Electrical
Function," Circulation Research, vol. 90, pp. 939-950, May 2002.
cited by other .
Smith, et al., "Electrical Alternans and Cardiac Electrical
Instabiltiy," Circulation, vol. 77 (1), pp. 110-121, Jan. 1988.
cited by other .
Smith, et al., "Subtle Alternating Electrocardiographic Morphology
as an Indicator of Decreased Cardiac Electrical Stability,"
Computers in Cardiology, pp. 109-112, 1985. cited by other .
Verrier, et al., "Risk Identification by Noninvasive Markers of
Cardiac Vulnerability," Foundations of Cardiac Arrhythmias-Basic
Concepts and Clinical Approaches, P. Spooner and M. Rosen (eds.),
Marcel Dekker, Inc., pp. 745-777, 2000. cited by other .
Verrier, et al., "Electrophysiologic Basis for T Wave Alternans as
an Index of Vulnerability to Ventricular Fibrillation," Journal of
Cardiovascular Electrophysiology, vol. 5, pp. 445-461, May 1994.
cited by other .
Walker, et al., "Repolarization alternans: implications for the
mechanism and prevention of sudden cardiac death," Abstract only,
Cardiovascular Research, vol. 57 (3), pp. 599-614, Mar. 2003. cited
by other .
Wayne, et al., "Exercise-induced ST segment alternans," Chest, vol.
83 (5), pp. 824-825, May 1983. cited by other.
|
Primary Examiner: Manuel; George
Attorney, Agent or Firm: Perkins Coie LLP
Claims
We claim:
1. A method of deriving spatial and/or temporal distributions of
action potentials across a heart, comprising: determining a T-wave
alternan waveform by (a) differencing a plurality of temporally
adjacent T-wave segments to obtain preliminary alternan waveforms,
(b) compensating for disturbances and/or ectopic beats in the
preliminary alternan waveforms, and (c) computing a final alternan
waveform; and decomposing the final alternan waveform to provide
information relating to changes in cardiac action potentials of the
subject, wherein decomposing the final alternan waveform comprises
fitting the final alternan waveform with a summation of a plurality
of alternan model curves, with individual alternan model curves
being associated with a specific phase of the cardiac action
potential and scaled in amplitude such that a difference between
the final alternan waveform and the summation is reduced.
2. The method of claim 1 wherein the cardiac action potential
comprise specific phases of depolarization, refractory and
repolarization phases.
3. The method of claim 1 wherein specific phases of the cardiac
action potential comprise depolarization, refractory and
repolarization phases.
4. The method of claim 3 wherein further comprising constructing
the alternan model curves by determining the difference between
augmented and diminished states of the associated alternating
action potential phase.
5. The method of claim 1 wherein amplitudes of scaling factors are
used in the assessment of cardiac instability.
6. A method of deriving spatial and/or temporal distributions of
action potentials across a heart, comprising: determining a T-wave
alternan waveform by (a) differencing a plurality of temporally
adjacent T-wave segments to obtain preliminary alternan waveforms,
(b) compensating for disturbances and/or ectopic beats in the
preliminary alternan waveforms, and (c) computing a final alternan
waveform; and decomposing the final alternan waveform to provide
information relating to changes in cardiac action potentials of the
subject, wherein the preliminary alternan waveforms associated with
a selected time period are used to assess the risk of cardiac
instability, and wherein the risk estimate includes a first measure
proportional to the simultaneous alternan voltage difference
between preliminary alternan waveforms associated with distinct
physiologic signals and a second measure inversely proportional to
the spatial separation of the regions of the subject's heart
monitored by the same distinct physiologic signals.
7. The method of claim 6 wherein the spatial separation of the
regions of the subject's heart monitored by the distinct
physiologic signals is estimated from the locations of a plurality
of monitoring electrodes placed on the subject.
8. A system for deriving spatial and/or temporal distributions of
action potentials across a heart, comprising: a data source
configured to obtain and/or retain data of a physiologic signal
having substantially repetitive waveforms of a heart beat; and a
computer operatively coupled to the data source, the computer
having a computer operable medium containing instructions for (1)
determining a T-wave alternan waveform by (a) differencing a
plurality of temporally adjacent T-wave segments to obtain
preliminary alternan waveforms, (b) compensating for disturbances
and/or ectopic beats in the preliminary alternan waveforms, and (c)
computing a final alternan waveform, and (2) decomposing the final
alternan waveform to provide information relating to changes in
cardiac action potentials of the subject, wherein the instructions
for decomposing the final alternan waveform comprise fitting the
final alternan waveform with a summation of a plurality of alternan
model curves, with individual alternan model curves being
associated with a specific phase of the cardiac action potential
and scaled in amplitude such that a difference between the final
alternan waveform and the summation is reduced.
9. The system of claim 8 wherein the instructions for decomposing
the final alternan waveform are related to specific phases of the
cardiac action potential comprise depolarization, refractory and
repolarization phases.
10. The system of claim 9 wherein the computer operable medium
further comprises instructions for constructing the alternan model
curves by determining the difference between augmented and
diminished states of the associated alternating action potential
phase.
11. The system of claim 10 wherein amplitudes of scaling factors
are used in the assessment of cardiac instability.
12. The system of claim 8 wherein the preliminary alternan
waveforms associated with a selected time period are used to assess
the risk of cardiac instability, and wherein the risk estimate
includes a first measure proportional to the simultaneous alternan
voltage difference between preliminary alternan waveforms
associated with distinct physiologic signals and a second measure
inversely proportional to the spatial separation of the regions of
the subject's heart monitored by the same distinct physiologic
signals.
13. The system of claim 12 wherein the spatial separation of the
regions of the subject's heart monitored by the distinct
physiologic signals is estimated from the locations of a plurality
of monitoring electrodes placed on the subject.
Description
TECHNICAL FIELD
The present invention relates to methods and apparatus for
determining T-wave alternan signatures (i.e., morphology and
polarity) derived from a physiologic signal representative of a
subject's heart activity; assessing changes in the myocardium
Action Potential ("AP") through analysis of the alternan signature
derived from a physiologic signal representative of a subject's
heart activity; and/or assessing spatial disassociation of alternan
characteristics that are likely associated with the initiation of
re-entrant arrhythmias.
BACKGROUND
T-wave alternans are characterized by a pattern of alternations in
the amplitude of the T-wave component of an ECG, where the even
beats systematically display a different amplitude than the odd
beats (an "ABABAB . . . " pattern of beat signatures). Many prior
research efforts have found correlations between the amplitude of
the T-wave alternans during periods of increased heart rates, and
sudden cardiac arrest or arrhythmias. Verrier and Cohen, in their
chapter "Risk Identification by Noninvasive Markers of Cardiac
Vulnerability" (Foundations of Cardiac Arrhythmias, Spooner and
Rosen editors, Marcel Dekker, Inc., 2000), provide an overview of
past research and describe a signal processing method for
determining the presence of microvolt level alternans. Summarily
stated, the ECG signal is evaluated to identify sequential data
points within the T-waves. The amplitude of these selected points,
from successive beats, forms pseudo time series that are next
subjected to Fourier analysis to create a power spectrum; the power
at the Nyquist frequency of this spectrum provides an estimate of
the energy of the beat-to-beat fluctuations in the amplitude of the
T-wave. The power spectra from successive individual spectra
associated with different offset times within the T-wave coda are
averaged to establish a composite power spectra, which is claimed
to be useful in assessing patient risk for sudden cardiac arrest or
arrhythmias. Clinical observations and trials have shown that
persons who exhibit T-wave alternans at relatively low heart rates,
i.e., .about.110 bpm, are at greater risk of developing fatal
arrhythmias than those who exhibit alternans at heart rates
approaching their maximum target heart rate. Both this publication,
as well as U.S. Pat. Nos. 4,802,491; 5,148,812; and 5,713,367
relating to this and related approaches are incorporated herein by
reference.
While the above-described method may be valuable for establishing
the gross existence and severity of T-wave alternans, the analysis
is limited to only the average amplitude of the alternan signal
across the entire T-wave signal. Clinical experience with
stratifying patient risk of sudden cardiac death based upon this
simplistic characterization of the alternan signal are typified by
a high rate of indeterminacy--typically as high as 30% of the
patient tests for alternans are indeterminate.
SUMMARY
The invention is broadly directed to cardiac assessments derived
through T-wave analysis. Additional information contained within
the alternan signal may yield important insight into the
electrophysiology of the myocardium, including parameters that
quantify the phase of the Action Potential ("AP") that is
exhibiting an alternating pattern and the degree of zonal
disassociation across the heart (i.e.: out of phase alternans
across the heart that may be the source trigger for re-entrant
arrhythmias). These additional data may lead to an improved method
for patient risk stratification and a lower indeterminate
threshold. Various features of the invention are directed to
methods for determining T-wave alternan signatures (i.e.,
morphology and polarity) derived from a physiologic signal
representative of a subject's heart activity; assessing changes in
the myocardium Action Potential ("AP") through analysis of the
alternan signature derived from a physiologic signal representative
of a subject's heart activity; and/or assessing spatial
disassociation of alternan characteristics that are likely
associated with the initiation of re-entrant arrhythmias.
As will be discussed in more detail below, some or all of these
features can be used to assess the cardiac condition of a subject.
In all embodiments, an estimated T-wave alternan signature for a
given heart rate is needed. This estimated T-wave alternan
signature includes derived waveform morphology (signal) while
preserving the polarity of the waveform, which provides data
heretofore unavailable by the prior art methods of cardiac
assessment through T-wave analysis. Robust embodiments include
multiple T-wave alternan estimates for multiple heart rates across
at least one signal source, such as at least one conventional ECG
Stress test lead.
In certain embodiments, a physiologic signal representative of a
subject's heart activity is acquired and the T-wave component of
selected heartbeats is identified. The T-wave components of
adjacent heartbeats are differenced to obtain a gross estimate of
resultant alternan signatures. The gross estimate is constructed to
include and preserve amplitude polarity information. At least one
and preferably several signal processing functions are performed to
derive at least one desired alternan signature estimate for the
selected heart beats, which is statistically correlated to and
representative of the alternan signature of the selected heart
beats.
The derived alternan estimate is preferably one of many such
estimates representative of various cardiac conditions induced by
stress testing the subject. A feature of an embodiment of the
invention relates to the reporting of the derived data. For
example, an embodiment of a reporting feature includes the
simultaneous visual display of a plurality of derived alternan
signature estimates in matrix form. In such an embodiment, the
plurality of derived alternan estimates are associated with a
corresponding plurality of heart rates by displaying temporally
adjacent estimates adjacent to one another. In this manner, an
analyst is readily able to discern changes in the alternan waveform
morphology over the range of heart rates being reported. Moreover,
the reporting feature can further include simultaneously displaying
visual representations, either numerically or graphically, of the
relative alternan waveform amplitudes derived from each physiologic
signal. For example, a plurality of alternan waveform estimates
derived from a plurality of ECG leads are presented in such a
format.
Another feature of several embodiments of the invention is to
normalize the acquired data to provide a better correlation between
the alternan estimate and the actual heart condition. Motion
artifacts, muscle artifacts, system noise, respiratory artifacts or
other noise present in at least one physiologic signal
representative of a subject's heart activity can obscure the
alternan estimate. To mitigate such noise, the acquired data is
normalized so that the alternan estimate more closely correlates to
the actual condition of the subject's heart. In one embodiment of
the normalization procedure, systemic amplitude fluctuations and
baseline wander in the waveform are characterized. The associated
effects on the signal are then minimized by correcting for
amplitude gain and DC bias to achieve a more accurate alternan
estimate for a plurality of repeating waveforms.
Several embodiments of the invention also increase the real-time
reporting ability of certain results and reduce random or
stationary noise by smoothing and sub-sampling the gross alternan
estimates. Such noise reduction can be achieved by calculating
median or average values and curve fitting using first or second
order polynomials. In a preferred embodiment, time domain segments
(time bins) of a given alternan estimate are established, which
preferably reduce the number of data points to about 15 to 25.
Suitable noise reduction algorithms, such as those described above,
are applied to each time bin, thereby yielding a smoothed estimate
of the alternan signature of interest. This method is then applied
to a suite of temporally adjacent alternan estimates until a
desired number of alternan estimates have been derived.
The usefulness of the smoothed alternan estimates can be enhanced
by obtaining a reference curve from these estimates, such as by
averaging the curves or preferably finding the median curve. From
this reference curve, a weighting factor can be established and
used to determined a weighted average alternan estimate of the
suite of smoothed alternan estimates derived above. In a specific
embodiment, the root mean square (RMS) of the difference between
the reference curve and each of the smoothed T-wave alternan
estimates from the suite of heartbeats is determined. Smoothed
alternan estimates that are similar to the reference curve, i.e.,
those wherein the RMS value is small, are weighted more heavily
than those that are dissimilar to the reference curve, i.e., those
wherein the RMS value is large. The derived weighting factor is
then applied to each alternan estimate and the weighted smoothed
estimates averaged to yield a robust alternan estimate for the
suite of heartbeats under consideration, or portions thereof.
Yet another feature of several embodiments of the invention manages
or otherwise compensates for disruptive events, such as premature
beats, pauses or other disruptions to a steady cardiac rhythm, that
may reverse the polarity of the alternan signature. Adjustment for
the presence of disruptive events is generally desirable for many
polarity sensitive embodiments of the invention. By monitoring the
polarity of each alternan signal within a suite of heartbeats,
adjustments to the polarity of alternan estimates following a
disruptive event can be applied.
Still another feature of a specific embodiment provides a basis for
associating certain types of alternan signatures with physiological
changes in the action potential (AP) of a subject's heart. It has
been found that a relationship exists between epicardial AP
alternations and T-wave alternans. For example, the three major
forms of epicardial AP alternations, i.e., depolarization,
refractory, and repolarization phases, are associated with three
distinct T-wave alternan signatures. An aspect of this feature is
to ascertain data from the T-wave alternan estimates that represent
specific characteristics of AP alternations. In one embodiment, at
least three model curves are established that represent the
alternation in ECG signal associated with alternation in each phase
of the AP. Through a simultaneous curve fitting method, the
estimated alternan signal is decomposed into components
representing the contribution from each of the three distinct AP
processes. Thus, by analyzing the waveform of a T-wave alternan
estimate, one is provided with information regarding the affected
phase of the epicardial AP. Risk estimates of cardiac instability
may be developed from these distinct estimates of AP
alternation.
Preserving the full waveform of the alternan signal, including
recording consistent amplitude polarity, supports an assessment of
cardiac alternan disassociation wherein distinct regions of the
heart display different alternan characteristics. These
out-of-phase alternan patterns establish voltages across regions of
the heart and may trigger arrhythmias. A risk assessment method can
then be developed that quantifies the severity of the alternan
disassociation based upon the simultaneous voltage differences of
the alternan signatures and the spatial separation of the regions
sampled by the distinct physiologic signals.
An embodiment of one method for determining T-wave alternan
signatures in accordance with the invention comprises: (1)
acquiring electrophysiological data (a beat series) from a
subject's heart of sufficient duration wherein such data includes
electrical signals corresponding to T-wave data found in an
electrocardiogram ("ECG"); (2) identifying T-wave segments within
the beat series data for use in the analysis that account for
ectopic beats and other significant changes that may disrupt the
alternan pattern; (3) correcting the data for baseline wander and
motion artifacts associated with respiration and other noise; (4)
differencing adjacent beats within the beat series while retaining
polarity and morphology information to compute initial estimates of
the alternan signature for the series; and (5) smoothing and
stacking the individual estimates to lower noise and provide a
robust estimate of the alternan signature for the beat series. This
method can further comprise (6) decomposing the alternan signature
into components related to changes in the depolarization,
refractory and repolarization components of the myocardium AP;
and/or (7) reporting the alternan signatures. In addition, optional
procedures can be employed to assess the severity of spatial
disassociation of the alternan signature in other embodiments of
methods in accordance with the invention.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1A is a flow chart illustrating a method for determining and
representing a T-wave alternan estimate in accordance with an
embodiment of the invention.
FIG. 1B is a diagram schematically illustrating stages of the
method for determining and representing an alternan estimate in
accordance with the embodiment of the invention shown in FIG.
1A.
FIG. 2A is a graph illustrating an ECG and the reference points
corresponding to activation and recovery of the Atria (P); the
ventricle activation phases Q, R and S, forming the QRS complex;
the recovery or re-polarization phase T of the ventricles; and the
R-R time interval between consecutive beats as measured between the
peaks of the R phase.
FIG. 2B is a graph showing the alignment of QRS complexes from a
cross-correlation.
FIG. 3 is a graph illustrating a sample ECG Beat Sequence and a
computed Median beat in accordance with an embodiment of the
invention.
FIG. 4 shows the alignment of the P-S signature of the average beat
(Ave(i)) with the beat sequence and the resulting amplitude and DC
offsets (G and C respectively) that mitigates the RMS error between
the average and the beat sequence in accordance with an embodiment
of the invention.
FIG. 5 is a graph illustrating the computed gain for a sequence of
beats that mitigates the amplitude and baseline wander associated
with the respiration signal contained in the ECG in accordance with
an embodiment of the invention.
FIG. 6 is a schematic illustration showing an example of
overlapping windows used to bin and smooth the alternan estimate
data in accordance with an embodiment of the invention.
FIG. 7 is a graph showing sixteen individual smoothed estimates of
alternan signature computed from 17 successive beats in accordance
with an embodiment of the invention.
FIG. 8A is a diagram showing the estimates of smoothed alternan
signatures computed from 16, 32 and 64 alternan estimates from
successive beats in accordance with an embodiment of the
invention.
FIG. 8B is a graph illustrating an ECG signal interrupted by
ectopic beats, dividing the entire sequence into multiple segments
of contiguous beats.
FIG. 8C is a diagram showing the polarity reversal of the smoothed
alternan estimate that may occur when an ectopic beat or other
disruption to the heart rhythm occurs.
FIG. 9 is a graphic overlay of APs and T-wave formation wherein the
T-wave represents the difference between the endocardial and
epicardial APs in accordance with an embodiment of the
invention.
FIG. 10 is a graph illustrating an alternan signal associated with
amplitude alternans in the depolarization phase of the epicardial
AP in accordance with an embodiment of the invention.
FIG. 11 is a graph illustrating an alternan signal associated with
alternans in the epicardial refractory period of the AP in
accordance with an embodiment of the invention.
FIG. 12 is a graph illustrating an alternan signal associated with
alternans in the epicardial repolarization phase the AP in
accordance with an embodiment of the invention.
FIG. 13 is a diagram illustrating the parametric form of three
curves used to decompose the alternan signature into the three
phases of the associated AP in accordance with an embodiment of the
invention.
FIG. 14 is a graph illustrating an example of color coding
corresponding to the amplitude of the alternan signature in
accordance with an embodiment of the invention.
FIG. 15 is a representation of a composite display of the alternan
signature for an entire stress test, including data for 8
independent ECG leads in accordance with an embodiment of the
invention.
FIG. 16 is a graph showing a possible display of average alternan
amplitude and heart rate for a stress test in accordance with an
embodiment of the invention.
DETAILED DESCRIPTION
The following discussion is presented to enable a person skilled in
the art to practice the invention. Various modifications to the
disclosed embodiments will be apparent to those skilled in the art,
and the generic principles herein may be applied to other
embodiments and applications without departing from the spirit and
scope of the present invention as defined by the appended claims.
Thus, the present invention is not intended to be limited to the
embodiments presented, but is to be accorded the widest scope
consistent with the principles and features disclosed herein.
FIG. 1A is a flow chart of a method 100 for determining and
representing a T-wave alternan estimate in accordance with an
embodiment of the invention, and FIG. 1B is a diagram schematically
illustrating the stages of the method 100 shown in FIG. 1A.
Referring to FIGS. 1A and 1B together, the method 100 includes a
first stage 102 comprising acquiring electrophysiological data
(e.g., a heartbeat series) of sufficient duration to include
electrical signals corresponding to T-wave data. The method 100
also includes a second stage 104 comprising identifying the T-wave
segments within the series of heartbeats. The second stage 104, for
example, can further include assessing the identified T-wave
segments to compensate for ectopic beats and other significant
changes that may disrupt the alternan pattern. Referring to FIG.
1B, the first stage 102 and second stage 104 are graphically
illustrated using an electrocardiogram (ECG) readout in which the
T-wave segment is identified for sequential heartbeats A and B by
bracketed arrows.
The method 100 continues with a third stage 106 that includes
correcting the acquired data for baseline wander and motion
artifacts caused by respiration, movement and other sources of
noise. The third stage 106 in FIG. 1B graphically illustrates
subsequent T-wave segments that have been corrected for baseline
wander and artifacts caused by such noise. After the third stage
106, the method includes a fourth stage 108 in which initial
estimates of the alternans between subsequent beats are calculated,
and a fifth stage 110 in which the estimated initial alternans are
smoothed and stacked to further lower the noise and provide a more
robust estimate of the alternan signature for a series of
heartbeats. The fourth stage 108 can include computing the initial
estimate of the alternans by calculating the differences between
adjacent T-wave segments within the beat series at common
respective time intervals in a manner that retains the polarity and
morphology information. The fifth stage 110 can further include (a)
providing a median estimate, (b) weighting the median estimate, and
then (c) determining a weighted average estimate of An
alternans.
The method 100 can optionally include a sixth stage 112 comprising
decomposing the alternan signature into components related to
changes in the (a) depolarization, (b) refractory, and (c)
repolarization components of the myocardium action potentials (AP).
The sixth stage 112 can accordingly determine different action
potential decompositions relative to the weighted alternan average
An to provide specific information to evaluate a patient as
described in more detail below. Several embodiments of the method
100 also include a seventh stage 114 in which the alternan
signatures for various electrodes are reported to an operator. As
shown in FIG. 1B, the seventh stage 114 can report the alternan
signatures in a graphical display illustrating the weighted average
alternan estimate An for a number of electrodes during a stress
test. An eight stage 116 further processes the data from stage 110,
for all leads and all time intervals, to derive a measure of
spatial disassociation of alternan voltages for use in assessing
patient risk from arrhythmias.
The embodiment of the method 100 shown in FIGS. 1A 1B is broken
down into discrete steps illustrating one embodiment of a series of
steps that achieve certain benefits of the invention. Other
embodiments of methods in accordance with the invention may not
include all of the stages 102 116 and still achieve several
benefits of the embodiment shown in FIGS. 1A B. For example, if
suitable digital ECG data is already available, it is not necessary
to create the data and certain steps need not be performed in order
to achieve the benefits of the invention. The following discussion
accordingly describes several of the stages 102 116 of the method
100 in greater detail with the understanding that the individual
stages can be eliminated or use other processes in other
embodiments of the invention.
A. Data Source and Protocol
The first stage 102 of the embodiment in this method 100 shown in
FIG. 1A involves acquiring electrophysiological data from a data
source using a suitable protocol. Clinically useful estimates of
low amplitude alternan signals preferably require a contiguous
heartbeat series of 64 128 beats with an approximately constant
heart rate. The number of contiguous beats may be in the range of
16 32 in the presence of low noise, and in exceptional conditions
could be as low as 2 beats. Because of the very low voltages being
analyzed, robust methods for reducing noise at all stages of data
acquisition and processing is highly desirable. While some sources
of noise cannot be reduced (see below), good ECG electrode
preparation such as with Quinton Cardiology Inc.'s QUICKPREP.RTM.
will significantly reduce the level of noise relating to the
electrode-subject interface. Details of this product and related
methods of use can be found in commonly owned U.S. Pat. No.
5,458,141, which is incorporated by reference. In addition or
alternatively, monitoring of electrode impedance and noise during
application will contribute to overall noise reduction.
Noise directly affects the length of the beat series used to
compute alternan estimates. Stationary noise not linked with or
created by the beat should decline approximately as the square root
of the number of beats included in the series--e.g.: 16 beats
should lower the noise by about a factor of 4, and 64 beats should
lower noise by about a factor of 8. A dynamic assessment of noise
conditions during data acquisition and processing can be used to
establish the number of beats necessary to provide reliable data.
Because ectopic beats and other cardiac events (e.g., a pause or a
rapid change in R-R interval) can disrupt the alternan pattern,
minimizing the number of consecutive beats required to obtain
reliable results affords more opportunity for diagnosis of subjects
with higher incidences of ectopic heart beats.
Noise can be generated from a multiplicity of sources and therefore
the identification and suppression, conditioning and/or filtering
of such noise is desirable. Common sources of subject-related noise
include motion artifacts (movement of the heart within the
pericardial cavity or chest movement induced by body impacts
encountered from walking or running during a treadmill stress
test); surrounding muscle contraction artifacts from body and arm
motion; breathing or respiration artifacts (both from the change in
chest impedance and the repositioning of the heart within the chest
as the lungs inflate); and, electrode-skin contact noise.
Some subject-related noise suppression and/or conditioning is
relatively easy to achieve and include actions such as having the
subject minimize and/or keep constant the rate of body impacts
during testing; maintaining a constant rate of physical exertion
(e.g., a constant pace of walking on a treadmill); attempting to
maintain a relatively constant respiration rate; etc. The
conditioning generally attempts to eliminate or stabilize the
frequency, duration, and/or amplitude of the noise, which
facilitates noise identification and removal. Thus, increasing the
rate of physical exertion during stress testing by having the
subject maintain a constant pace while increasing the elevation of
a treadmill is one means for conditioning the noise for later
filtering (and thus suppressing the effect of that noise). A
derivative of this approach is to monitor the heart rate and
dynamically adjust the grade of the treadmill to maintain the
target rate, thereby eliminating the variables associated with a
non-stable heart rate during the data acquisition phase.
In addition to external sources of noise, there is
clinical/laboratory data that suggests ectopic beats and other
cardiac events may re-set the alternan signal (i.e., the alternan
signal may shift a beat, switching from an even-odd pattern to an
odd-even pattern, thus disrupting the analysis). Therefore,
contiguous beat sequences should be selected for analysis that
exclude ectopic beats or other disruptions. For subjects with a
high ectopic rate it is necessary to extend the analysis by
combining multiple shorter beat sequences and using correlation
methods to determine the possible polarity change for each sequence
associated with a changing beat pattern, as discussed below.
A preferred stress profile to induce an alternan response is to
place a subject on a controllable, variable speed treadmill with a
variable incline feature such as the Q-Stress.RTM. cardiac stress
testing system manufactured by Quinton Cardiology, Inc. The data
acquisition protocol is preferably similar to a standard Burke
protocol, beginning with the subject at rest and recording a suite
of heartbeats for about 2 3 minutes. Next, the subject's heart rate
is increased by increasing the treadmill grade and then held
constant for about 2 3 minutes, during which time another suite of
heartbeats is recorded. This progression continues until the
desired maximum heart rate is achieved or the subject is exhausted.
Maintaining a constant treadmill speed, and thus a constant pace,
while increasing the grade stabilizes noise associated with body
motion artifacts and improves the estimate of alternan signature
associated with increasing heart rate. Alternatively, but with
somewhat higher motion noise associated with increasing heart rate,
other protocols such as the Bruce protocol could be used to
exercise the subject, where both the speed and grade are adjusted
periodically until the subject either reaches the desired maximum
heart rate or is exhausted.
B. Cross-Correlation and T-Wave Selection
After successfully recording appropriate ECG samples or retrieving
such samples from stored data at each desired heart rate, the
method 100 continues with the second stage 104 by ascertaining the
alternan component between temporally adjacent T-waves. One aspect
of this stage is consistently selecting from beat to beat the onset
time of the T-wave segment. The amplitude of the T-wave is very
large compared to the alternan heartbeats signal. In general, the
T-wave may have an amplitude of 300 1000 microvolts; the alternan
signal computed from the difference of adjacent T-waves may have an
amplitude of only a few microvolts. A mis-alignment of the T-wave
between two adjacent beats of 2.0 milliseconds (the standard
sampling rate for stress testing) can produce a false amplitude
anomaly of 5 10 microvolts. Therefore, it is important to use a
high sampling rate, e.g., 0.5 1.0 millisecond, and to accurately
align the T-waves in order to measure the signal and not be
overwhelmed by processing noise associated with mis-alignment of
the T-waves from beat to beat.
FIG. 2A illustrates an example ECG with the key phases identified.
The normal heart beat starts in the upper chambers of the heart
(atria) and the initial ECG phase that records this activation is
termed the P-wave; the bracket indicates the duration of the
P-wave. Following the activation of the atria the blood moves into
the lower chambers of the heart (ventricles) and activation of the
ventricle muscle both pumps the blood to the body and generates the
ECG phases Q, R and S, often referred to as the QRS complex.
Finally, the ventricle muscles recover (repolarize) in anticipation
of the next beat, creating the T-wave signal on the ECG. The time
interval between adjacent beats is generally measured between the
peaks of the R-wave and is referred to as the R--R interval. The
letter designations are commonly used to also specify specific
segments of the ECG. For instance, the PQ interval would be the
segment that begins with the onset of the P-wave and concludes with
the Q-wave.
FIG. 2B illustrates an example of aligned waveforms from a series
of heartbeats. The waveforms have a QRS complex 202 with an R-wave
peak 204. For a sequence of beats to be analyzed for an alternan
signal, the following steps are preferably used to select the
T-wave segment within each beat in a beat series. 1. Using the
R-wave peak 204 within the QRS complex 202 to approximately time
align each beat, compute a median beat estimate for the selected
sequence of beats for at least one, and preferably all, ECG leads.
2. Window the QRS segment from the median beat estimate computed in
step 1 and cross-correlate this windowed QRS pulse with the ECG
data, finding the peaks in the cross-correlation. Preferably, the
cross correlation metric should be based upon the peak of the sum
of the cross correlations across leads I, II, and V1 V6--i.e.: the
selected QRS correlation time point should be the same across all
leads. 3. Use the refined QRS onset time from step 2 to re-stack
the data, using either an average or median stack, thus developing
an improved estimate of the QRS complex. 4. Using the improved QRS
estimate, repeat steps 2 and 3 above, resulting in a final estimate
of QRS onset time for all beats in the sequence and all traces and
a best estimate of the QRS complex morphology. Save the QRS complex
for further use in computations discussed below. 5. The T-wave 210
for each lead is preferably selected based on the following
parameters: a. The onset time should be a few samples beyond the
maximum negative excursion of the S-wave 212 (and the same across
all leads). b. The duration of the T-wave, and hence the end time,
is more complex. Preferably, the analysis uses the entire T-wave,
but it is helpful to maintain a constant T-wave window length over
the entire test analysis. As heart rate increases, the P-wave for
the next beat may start to ride on the end of the T-wave from the
previous beat, adding noise to the analysis. The target heart rate
and the associated estimate of the duration of the R--R interval
(D.sub.R-R) at peak exercise, along with the duration of the
interval from the onset of the P wave to the end of the S phase
(D.sub.P-S), should be used to compute a maximum window length for
the T-wave (Maximum Length=D.sub.R-R-D.sub.P-S. The T-wave window
length should be computed from the initial (resting) data and held
constant for the entire test. Preferably, the end of the windowed
T-wave should be well beyond the peak of the T-wave--as discussed
below, most of the alternan signal will be in the segment between
the end of the S-wave and the peak of the T-wave.
The cross-correlation times of the QRS complex also form the basis
for tracking R--R intervals and associated dispersion, and for
identifying anomalous pauses or jumps in heart rate that may re-set
the alternan signal. The cross-correlation times are useful in this
analysis and are generally retained for subsequent use.
C. T-wave Normalization for Baseline Wander
The third stage 106 of the method 100 processes the data to
mitigate the affects of noise. The ECG data is influenced by many
sources of noise, including high frequency muscle artifact and
system noise as well as long period noise associated with
respiration and body movements. Referring to FIG. 3, the raw ECG
data typically includes both a baseline wander and amplitude
variations associated with respiration. In this figure the
amplitude variations can be seen by examining the difference
between the peak of the R-wave and the trough of the S-wave; this
difference is smallest for the beats on the left and right side of
the figure and maximum for the beats in the central portion of the
ECG. This figure also illustrates baseline wander as detected by
observing how the onset of the QRS complex rises for the central
beats and falls for the beats on the left and right sides of the
ECG. Even though the ECG data includes such baseline wander and
amplitude variations, a robust estimate of the average or median
beat, Ave(i), can be computed, as discussed in the previous
section, and is shown adjacent to the ECG trace.
The average or median beat Ave(i) estimate can be compared with the
individual beats to derive an amplification gain factor and a DC
shift factor. For example, the median or average beat, Ave(i), can
be scaled by an amplification factor G(m) and DC shift factor C(m)
to minimize the least square difference with each beat in the
sequence. To mitigate or prevent introducing systematic bias or
noise into the T-wave portion of the signal, the minimization
window should focus on just the P-S beat segment and solve for the
optimal G and C for each beat, as illustrated in FIG. 4.
The system of equations to be solved for each beat are:
G.times.Ave(i)+C=ECG(i) In matrix notation the least squares
solution to this system of equations is:
.times..function..times..function..times..function..times..times..functio-
n..times..function..times..function. ##EQU00001## where n is the
number of data points in the P-S interval. Solving for the inverse
yields:
.times..times..function..times..function..times..function..times..times..-
function..times..times..function..times..function. ##EQU00002##
.times..function..times..times..function..times..function..times..times..-
function..times..function..times..times..function..times..function.
##EQU00002.2##
Once G and C are derived for each beat, a cubic polynomial function
(F.sub.C and F.sub.G) can be computed that smoothly connect four
values of G or C associated with four consecutive beats. The
computed middle segment of the resulting function, between the
second and third values for G or C.sub.1 is used to correct for
gain and DC bias for the T-wave in this segment, following:
.function..function..function..function. ##EQU00003##
As noted by many previous investigators (see, e.g., Moody et al.,
"Clinical Validation of the ECG-Derived Respiration (EDR)
Technique," Computers in Cardiology, p 507 510, 1986, which is
incorporated by reference), the ECG signal is modulated by
respiration. The amplitude variations are caused by mechanical
movement of the electrodes relative to the heart, rotation of the
heart within the chest, and changes in chest impedance as the
patient breaths. The herein computed gain for each beat can also be
used to derive the respiration rate. The curve shown in FIG. 5 is
the computed gain for an example ECG, the numbers along the X-Axis
are the beat sequence numbers.
The instantaneous respiration rate can be computed by measuring the
time between peaks (e.g.: 7 seconds/breath) or by averaging the
rate over longer time periods. Alternatively, the peak in the
Fourier transform of this series (i.e.: the series constructed from
the consecutive gain correction values for each beat) provides an
estimate of the respiration rate. The energy at the Nyquist
frequency also provides an estimate of the respiration noise that
has an alternans rate. These derived estimates of respiration rate
are preferably included in the noise analysis and estimation of
alternans reliability.
D. T-wave Alternan Estimate
After extracting the T-waves from a contiguous suite of non-ectopic
beats in the second stage 102 and processing the data to compensate
for baseline wander and systematic amplitude variations in the
third stage 104, the waveform difference between the adjacent beats
is next computed in the fourth stage 108. This difference in
waveforms for adjacent heart beats provides an initial estimate of
the T-wave alternans. More specifically, T-wave alternans are
characterized by an amplitude of the T-wave that is alternating
every other beat. For example, in the beat sequence 1, 2, 3, 4, 5,
. . . the even beats would have an amplitude augmentation relative
to the odd beats. Hence, estimates of the alternans are computed
through difference of the even beats minus the odd beats. For a
sequence of m beats, 1, 2, 3, 4, 5 . . . the alternans estimates
are: (2-1), (2-3), (4-3), (4-5) and so on. Re-ordering, this is:
(2-1), -(3-2), (4-3), -(5-4) and so on. So, the jth estimate of the
alternans, at time position i in the T-wave, can be computed from
the normalized T-waves from each beat by:
Alternan(i,j)=(-1).sup.j(T(i,j)-T(i,j-1)) E. Smoothing and
Sub-Sampling the Alternan Estimates
The method 100 continues with the fifth stage 110 by smoothing,
sub-sampling, and further refining the alternan estimates. The
above computed estimates of the alternans will typically contain
about 600 700 data points (at 2000 samples per second). The
alternan signal has a somewhat longer period, relative to the 1000
Hz Nyquest frequency of the raw ECG data, and smoothing of each
individual alternan estimate is an effective way to further reduce
random or non-stationary noise. Computationally, it is preferable
to reduce the number of data points that are used in the subsequent
computations to a number in the range of 15 25 (an odd number being
desired).
Referring to FIG. 6, the alternan estimate computed over the
duration of the windowed T-wave can be divided into bins of data. A
simple median or average over the specified time bins usually
provides sufficient smoothing. However, a first or second order
polynomial may also be fit through the data and the mid-point of
the fitted curve used as the average value for the bin. The bins
should be overlapping and follow the general structure illustrated
in FIG. 6.
This procedure reduces the estimate of the alternans to around 21
points that spread uniformly over the duration of the alternan
signal. This also improves the signal to noise ratio by
approximately a factor of 5. Clearly, the number of bins and the
bin lengths can be adjusted as appropriate for the length of the
alternan estimate. In general, an odd number of bins in the range
of 15 25 provides acceptable smoothing while retaining the complex
morphology of the alternan signal. The smoothed alternan estimate
is designated as Alt_Smoothed(i,j), where i is the time position
ranging from 1 to about 21 for the jth alternan estimate.
The plot in FIG. 7 illustrates 16 individual smoothed alternan
estimates, Alt_Smoothed(i,j=1,16), smoothed from an initial 160
points. Although each smoothed alternan estimate still displays
noise, the general fabric of the alternan signal is beginning to
emerge in the 16 independent estimates.
F. Median Estimate of the Alternans
The fifth stage 110 can further include determining a median
estimate of the alternans over a period of several heart beats. The
above discussed processing can be computed for a contiguous suite
of beats, resulting in a suite of smoothed estimates of the
alternan signal. In general, the signal to noise level from any
single estimate will still be quite low and ensemble averaging of
many estimates, perhaps as high as 128 depending upon noise
conditions, may be necessary. It is also common to have occasional
noise bursts that are localized in time, e.g., a muscle artifact
spike, causing estimates computed from the affected beat to be
exceptionally noisy. The exceptionally noisy estimates resulting
from noise spikes can be identified and suppressed, and/or
prevented from significantly lowering the signal to noise
enhancement that should be obtained from normal signal averaging.
This can be accomplished by a two step process. First, an
approximate estimate of the alternan signal is developed, and then
a weighted average computation is performed based upon the root
mean square (RMS) difference between the estimate for the suite and
an individual estimate. The estimate for the suite preferably uses
a median estimate, Median(i), as it is robust in the presence of
occasional noise bursts.
G. Weighting Estimates
After determining the median estimate of the alternans, the fifth
stage 110 can continue by weighting the individual smoothed
alternan estimates. The median estimate established previously may
not be useful by itself; however, it finds utility with respect to
establishing a weighting factor for each of the smoothed alternan
estimates. This weighting factor is preferably used in averaging
the individual smoothed alternan estimates. To establish the
weighting factor, the RMS of the difference between the median
estimate, Median(i), and each of the individual smoothed alternan
estimates, Alt_Smoothed(i,j), are computed. When the RMS is large
(i.e. the smoothed alternan estimate is substantially different
from the median estimate), the weighting factor for the alternan
estimate is low, and vice versa. The weighting factor for the jth
alternan estimate is defined as:
.function..times..times..function..times. ##EQU00004## The
weighting factors may be adjusted such that 5 10% of the alternan
estimates that best fit the median estimate are uniformly weighted;
strict adherence to the weighting equation could lead to an
exceptional weight for the chance case where an estimate nearly
exactly equals the median. H. Weighted Average Estimate of the
Alternan Signal
The fifth stage 110 can further include determining a weighted
average estimated alternan An. Using the smoothed alternan
estimates and the associated weighting factors, the weighted best
estimate for the smoothed alternan signal is:
.function..times..times..times..function..times..function.
##EQU00005##
The Standard Deviation should be computed from:
.times..function..times..times..function..times..times..times..function..-
times. ##EQU00006## for m estimates of the alternan signal at n
values for each individual estimate. Weight(j) is defined above.
The industry standard gross estimate of the alternan amplitude may
be computed from An(i) by:
.times..times..times..function. ##EQU00007##
The weighted stack can be computed over any number of estimates of
the alternan signal. FIG. 8A shows computations from 16, 32 and 64
estimates of the alternan signal. A representative example of
dispersion of the individual estimates is shown in the previous
FIG. 7 for the 16 estimated averages (labeled "Mode-16").
I. Ectopic Beat Management
Ectopic beats and other disruptions to a steady rhythm may cause a
reset of the alternan signal, potentially changing from an
odd-even-odd pattern to an even-odd-even pattern. FIG. 8B
illustrates a beat sequence interrupted by three ectopic beats that
segments the overall ECG into subsections of contiguous beats,
marked as sequences A, B and C. The impact of a reset is to change
the sign or polarity of the alternan signal as illustrated in FIG.
8C, i.e.: the morphology is "up-side-down" relative to the
preceding pattern. As heart rate is the primary driver for
triggering alternans, a reasonable assumption is that the
underlying AP biophysics is unchanged by the event and the
morphology, but not necessarily the polarity, is approximately
stationary. This leads to a method to join together multiple
shorter continuous beat sequences that have been interrupted. As
suggested by FIG. 8C, the shape of the alternan signal can be used
to determine if the polarity has changed after a disruptive event.
If the cross correlation of the alternan estimates between the
sequences before and after a disruptive event is greater than the
cross correlation when one of the alternan estimates is reversed in
polarity then no disruption has occurred. Algorithmically, if
.times..times..function..times..function.
.times..times..function..times..function. ##EQU00008## (where k is
the lead number) is true then a polarity reversal has not occurred
and the sequence of beats following the disruption can be used
without correction. If this expression is false then a reversal has
occurred and the polarity of the subsequent alternan estimates
should be reversed for all leads after the disruptive event. This
method can be used to improve signal to noise provided a rough
estimate of the alternan morphology is emerging in the sequence in
some of the leads. This method is employed by selecting contiguous
beat sequences in stage 102, for instance segments A, B and C in
FIG. 8B, that are not interrupted by ectopic beats or other
disruptions, processing each sequence through stage 110, applying
the above criteria to the weighted average alternan estimate
computed for the first two sequences (A & B), correcting the
second sequence if necessary for polarity reversal, and computing
the ensemble weighted average for the two sequences. Using this
combined estimate of the alternan signature, the polarity of the
alternan estimate for the next sequence (e.g.: sequence C in FIG.
8B) can be assessed, corrected if necessary and combined with the
estimate from the combined first two estimates. This method can be
continued to each sequence in the selected ECG being processed in
stage 102, thus increasing the overall signal to noise ratio of the
alternan estimate. Short sequences may contain high noise that
renders them of questionable value for inclusion in this process.
The standard deviation estimate computed in stage 10, or other
estimates of noise, may be used to decide when to exclude a
sequence. J. Best Fitting Model
In several embodiments, the method 100 also includes the sixth
stage 112 of decomposing the alternans into components that can be
correlated to specific conditions and specific areas. Before the
present invention, the prior efforts to ascertain information from
T-wave alternans looked only at changes in alternan amplitude, and
this was done primarily over a narrow range of heart rates.
However, it has been ascertained that T-wave analysis yields
valuable information beyond simple amplitude related information.
For some changes in the myocardium, the T-wave alternan signal will
actually increase in some segments and decrease in others--the
alternan signal is not just variability in peak amplitude, but
includes changes in shape. It is therefore desirable to provide a
means for associating the best estimate of the alternan signal (An)
into an estimate of the nature of the AP alternation within the
myocardium and an associated measure of uncertainty. To make this
computation, several models for the T-wave alternan signals are
provided and a procedure has been developed to systematically
assess which model best fits the observations. This procedure
associates features within the alternan signal with distinctly
different phases of the myocardium Action Potential.
1. Action Potentials
The T-wave shape is strongly controlled by the shape of the APs of
the cardiac tissue. The relationship between the AP and the
observed surface ECG is complex--the AP may vary across the heart,
and the observed surface ECG results from the spatial/temporal
derivative of the distribution of potentials and activation times.
Nevertheless, a useful approximation, particularly for the V leads,
is that the shape of the T-wave is controlled by the difference
between the Endo- and Epicardium APs. Referring to FIG. 9,
activation of the myocardium begins at the Purkinje fibers within
the Endocardium 910 and propagates outwardly activating the
Epicardium 920 last. The QRS complex 930 results from the timing
differences in the activation across the myocardium and the T-wave
is controlled by the difference in the repolarization segments of
the APs.
Many studies have also documented that the Epicardium AP is most
sensitive to ischemic change, and lab studies have strongly linked
alternans with the onset of ischemia, while the Endocardium AP is
relatively stable to ischemic changes (see Ionic Current Basis of
Electrocardiographic waveforms, K. Gima & Y. Rudy, Circulation
Research, p. 889 896, 2002, which is incorporated by reference).
Thus, alternans most likely represent alternation of the Epicardium
AP.
Still referring to FIG. 9, the earliest part of the AP is the
depolarization phase 940 that exhibits a very abrupt onset and the
key driver for the QRS complex. The following plateau 950 is the
refractory period when the cardiac tissue is unable to respond to
additional stimulus. Finally, the myocytes re-polarize 960 in
preparation for the next cycle and the AP decays back to the
starting potential. The T-wave reflects the re-polarization process
and the peak of the T-wave corresponds to end of the Epicardium AP.
The importance of this observation is that variability in AP
morphology measured in lab specimens can be used as a guide for
understanding the range of likely variability in the T-wave
alternan signals. See "Mechanism Linking T-wave Alternans to the
Genesis of Cardiac Fibrillation," Pastore et al, Circulation, p.
1358, 1999, which is incorporated by reference. The following
paragraphs highlight several observed variations in AP shape and
the expected T-wave changes, and resulting alternan signals.
2. Pattern I--Variability in the Amplitude of Depolarization
One variation comes from changes in the strength or amplitude of
the depolarization phase, while the refractory and re-polarization
phases retain their characteristic shapes and time constants. FIG.
10 shows the AP, T-waves and alternan signal for this case,
illustrating variations between a strong Epicardium AP 1010 and an
amplitude diminished Epicardium AP 1020. The resulting T-wave
variability will alternate with an amplitude scaling directly
related to the variability of the Epicardium depolarization
amplitude. Note that the polarity of the alternan signal, i.e., a
maximum or a minimum, is dependent upon which beat--even or
odd--contains the alternan signal.
This is the most simple example, but clearly links a possible
alternan signal with alternating S-T segment elevation/depression,
commonly associated with ischemia and myocardial infarcts, that has
maximum amplitude at the end of the QRS complex--the S-T
Junction--and tapers to zero at the apex of the T-wave. This model
has an important auxiliary prediction: the alternation in
depolarization amplitude should also cause an alternation in the
QRS amplitude, which has rarely been observed with standard ECG
recordings. However, the extremely high frequency content of the
depolarization (.about.1000 Hz) is well above the band-pass of
common ECG equipment.
3. Pattern II--Variability in the Refractory Period
This model assumes that the depolarization amplitude of the
Epicardium and the time constants for re-polarization remain
constant, but the plateau refractory period oscillates in an
alternating pattern. FIG. 11 shows the associated APs, the T-waves
and the alternating duration of the plateau from long 1110 to short
1120. For this model the alternan signal will peak around the
mid-point between the end of the QRS complex and the maximum of the
T-wave, tapering to zero at both ends.
4. Pattern III--Variability in the Repolarization Time Constant
The third possibility is that the Epicardium repolarization time
constant alternates between beats. FIG. 12 shows one possible
variant in repolarization, with one phase exhibiting a rapid or
steep repolarization 1210 and the other a more modest slope 1220,
along with the change in the T-wave shape and the resulting
predicted alternan signal. Note that this alternan shape is
distinct from the previous two patterns, exhibiting a biphasic
character and tapering to zero at the end of the QRS complex and
the peak of the T-wave.
5. AP Variant Discussion
There are three key electrical activities that characterize the
shape of the cardiac AP: a depolarization causing an abrupt rise in
potential; a refractory plateau characterized by a slowly varying
potential; and, a repolarization with the rapid decay of the
potential and return of the heart to a state of excitability. Three
possible AP variants have been investigated as models that capture
the key observations reported in lab studies of measured APs.
The characteristic variations in each of the three phases of the
cardiac AP are predicted to be associated with three very different
alternan signatures. This suggests that the shape of the alternan
curve may lead to a diagnostic method for identifying and focusing
attention on specific cellular activities that are under duress in
the stressed heart. Alternan methods of the prior art that just
focus on the absolute average amplitude of the T-wave difference
ignore most of the potential data contained in the alternating
morphology. In addition, many AP studies have indicated that the
stressed heart can disassociate from uniform alternan behavior to
zones of tissue responding with different, or out-of-phase,
alternans, leading to significant electrical instabilities that
trigger re-entry and life threatening arrhythmias. Using the
alternan morphology and polarity information derived from different
surface electrodes in a standard 12 lead stress test offers a
promise of identification of alternan disassociation and improved
patient risk stratification.
It is important to note that the family of possible AP variability
is large and the above discussion is not meant to represent the
entire family of useful curves. Ongoing clinical studies will guide
the refinement and selection of curves that represent typical
observations. However, the general shape of the above three curves
are sufficient to support early clinical studies.
6. Matching Model Curves and Data
The above discussed model curves are part of a family of orthogonal
curves that can be fit to the best estimate of the alternan signal,
An(i), defined above. The following parametric curves capture the
morphology of the above discussed models and form a reasonable
starting point:
.function..times..function..pi.I ##EQU00009##
.function..times..function..times..pi.I.pi. ##EQU00009.2##
.function..times..function..pi.I.times..function..times..pi.I
##EQU00009.3## where i is the sample number and l is the number of
samples in the windowed T-wave between the start of the window and
the peak of the T-wave. The curves should be padded with zeros
between the peak of the T-wave and the end of the T-wave window.
The curves have been normalized to a peak to peak amplitude of 1.
The graphical form of these curves is shown in FIG. 13. After
computing these three curves, each should be smoothed and
sub-sampled using the same filter methods used above to smooth the
alternan estimates.
The alternan signal can be decomposed into components representing
the contribution from each of these distinct curves and AP
processes. This is done by minimizing the least squares error
between the model and the data by finding the optimal values for
A.sub.n and C in the equation:
A.sub.iP.sub.1(i)+A.sub.2P.sub.2(i)+A.sub.3P.sub.3(i)+C=An(i) where
each A.sub.n represents the amplitude of the corresponding model
curve and C represents any residual DC bias in the alternan
estimate. In a matrix notation this defines an over determined
system of equations:
.function..function..function..function..function..function..function..fu-
nction..function..times..function..function..function. ##EQU00010##
And the solution is:
.times..function..times..function..times..function..times..function..time-
s..function..times..function..times..function..times..function..times..fun-
ction..times..function..times..function..times..function..times..function.-
.times..function..times..function..times..function..times..function..times-
..function..times..function..times..function..times..function..times..time-
s..function..times..function..times..function..times..function..times..fun-
ction..times..function..times..function. ##EQU00011## K. Model
Standard Deviation
The weighted Standard Deviation should be computed from:
.times..function..times..times..function..times..times..times..times..fun-
ction..times..function..times..function..times. ##EQU00012## for m
estimates of the alternan signal at n values for each individual
estimate. Weight(j) is defined above. L. Reporting the Results
The above analysis will result in a very large amount of data. Good
reporting metrics and tools for visualizing the results and
efficiently communicating the clinical significance is considered
important. The following sections describe these areas.
1. Onset of Alternans and Disassociation
As best illustrated in FIG. 14, a color coding scheme of the
display follows the amplitude curve for each individual estimate of
the alternan signal. This permits easy visual assessment of
amplitude in addition to convenient evaluation of the alternan
signal signature. The complete test summary, for all leads, is
developed by compositing together each individual alternan estimate
as shown in FIG. 15. This is the most important summary graph that
forms the basis for clinical analysis. It has been designed to
clearly show the onset and amplitude of any statistically
significant alternan signal and highlight alternan disassociation
observed across the lead set, which should be visible as both
changes in shape from lead to lead and changes in color
(amplitude). Key elements of the display are: Time Scale--Left
Side: The test is graphically portrayed as a series of T-wave
alternan estimates during the course of the stress test. In this
example, the test was 18 minutes in length. T-wave alternan
estimates are computed from a sliding window of individual alternan
estimates. Leads--Top: For a standard 12-lead test the results for
the eight independent leads are shown. Leads III and the augmented
leads could be added if clinical needs dictate--but the selected
leads are expected to be sufficient for most applications. For
higher lead tests, such as a 15-lead test, additional lead results
may be added to the display. Heart Rate--Right Side: The computed
heart rate at intervals during the test are shown. The heart rates
are associated with the displayed estimates of the alternan signal.
Alternans: The curves shown on each lead panel are the smoothed
alternan estimate An(i) correctly placed vertically with regard to
the test time and heart rate. Color: The color spectrum scale may
be either dynamically scaled for the range of the alternan values,
or fixed to a constant color scale to facilitate comparisons
between different subjects. The spectrum should be 32 64 colors
deep. Color coding may be set to white or no-fill if the Standard
Deviation estimate for the alternan is greater than the maximum
amplitude (Amp) associated with the smoothed alternan estimate
An(i). Average Beats: The resting average beats may be shown at the
bottom of each lead column. It may also be useful to display the
average beats corresponding to the maximum heart rate or maximum
alternan signal.
2. Alternan Disassociation Index
Clinical studies have shown that alternans can disassociate or
become out of phase across the heart (i.e., one zone may exhibit a
high-low or odd-even pattern while the adjacent zone is exhibiting
a low-high or even-odd pattern). This reflects out-of-phase
Epicardium AP augmentations and diminutions across small spatial
zones (see, for example, "Mechanism Linking T-wave Alternans to the
Genesis of Cardiac Fibrillation," Pastore et al, Circulation, p.
1358, 1999, which is incorporated by reference), creating
significant electrical gradients and potentially triggering
re-entry and arrhythmia. The alternan induced electric gradient is
controlled by the voltage differences of the alternan signals, as
recorded by each lead, and the spatial separation on the heart
associated with the region of the heart sampled by each lead. This
leads to a metric or index for judging the severity of
disassociation useful in risk stratification:
.times..function..function..function..noteq. ##EQU00013## where j
and k represent the V lead index, from 1 6 for a standard 12-lead
test, for An(i) at time T in the test. This expression can be
generalized for higher lead tests.
3. Alternan Amplitude and Heart Rate Trending
FIG. 16 presents selected data regarding the averaged alternan
signal for a given time period during the stress test. The
horizontal axis represents the stress test duration. The left axis
is the alternan amplitude and the right axis is the heart rate.
Preferably, the alternan data for each desired lead is plotted as
shown.
* * * * *